A literature review in the Elsevier Journal of Petroleum Science and Engineering by Timur Bikmukhametov and Johannes Jäschke (NUST, Trondheim) compares first principles and machine learning based approaches to virtual flow metering (VFM). VFM can replace expensive hardware metering devices and reduce the need for well tests. The review compares traditional VFM methods and commercial software with emerging data-driven methods.
Data-driven approaches circumvent the simulation of complex systems or processes for which the exact solution can be difficult to find numerically. The review covers operational experience from trials of data-driven models. Much commercial software (OLGA, K-Spice/LedaFlow, FlowManager and others) fall in the ‘first principles’ camp.
A notable exception to this is FieldWare Production Universe (FW PU), a data-driven VFM software package used by Shell. FW PU is said to be robust and accurate for conventional and multizone wells and capable of tracking changes in production regimes. Baker Hughes’ NeuraFlow is also cited as being data-driven and working for extended periods without re-calibration. The authors discuss, with copious references, different approaches to solving the ‘highly non-linear problems’ involved in data-driven VFM, notably the computationally efficient ensemble Kalman filtering (EnKF).
Summing up, today’s commercial VFMs are deployed both as a standalone solution or as a back-up system for physical multiphase flow meters. However, in these systems, model and PVT data tuning and transient flow behavior are problematical. Data-driven methods ‘are becoming popular as more field data is available and as algorithms advance and compute power increases’. The authors consider that a promising research direction is the development of a hybrid VFM that combines both first principles and data-driven modeling. The study was carried out as a part of SUBPRO, a NNTU unit.
Another Journal of Petroleum Science and Engineering paper, ‘Leveraging digital rock physics workflows in unconventional petrophysics: A review of opportunities, challenges, and benchmarking’ by Ayaz Mehmani, Carlos Torres-Verdín (both UTx Austin) and Shaina Kelly (ConocoPhillips) reviews the state-of-the-art in numerical simulation and pore-network modeling. The authors observe that ‘despite advances in micro-computed tomography (µCT) and scanning electron microscopy (SEM) techniques, obtaining sufficient information to capture dual-scale porosity and surface textures remains ‘a formidable challenge’. The paper reviews the current status of digital rock physics (DRP) in tight and/or diagenetically-altered rocks.
The review concludes that performing DRP on a core scan is not enough, even for many conventional rocks. It is crucial to interface µCT scans with experimental data from core analyses or microfluidics. Most current studies of shale tend to focus on discrete pore-scale scenarios that are often not benchmarked with SCAL data. The paper offers a technical roadmap for the ‘robust application of unconventional DRP for the petrophysical and general subsurface community’.
An enthusiastic report from Norway’s Norce R&D body has it that novel approaches to understanding the reservoir have ‘triggered large volumes’ of oil resources. The Norce-developed methods use the ensemble Kalman filter (EnKF) to merge large amounts of diverse data on the reservoir.
Norce researcher Geir Evensen manages the RCN Petromaks-2 project DIGIRES where the EnKF technology is being developed for operational use in ‘digital decision support’. The project is a collaboration with seven oil companies. EnKF, originally used weather forecasting is now said to work well in combination with reservoir modeling and simulation. Several oil companies, including Equinor, are now using EnKF technology in their recommended workflow.
A Rystad Energy presentation at the 2020 Offshore Strategy Conference in Stavanger claimed that ‘reservoir characterization, including the use of EnKF technology, is estimated to account for over half of the technology area’s realized reserve increase of 540 million barrels of oil equivalents’. Make of that what you will! More from the Norce ‘Digifuture’ minisite.
A presentation at the 2019 Norwegian Conference on ICT by researchers at Sirius (an Oslo University unit) investigated the use of ‘rewriting logic’ to formalize the description of geological processes. The presentation on ‘Geological multi-scenario reasoning’ has it that the approach may replace today’s ‘ad hoc manual work practices’ for developing and communicating multiple geological hypotheses. The formal framework for geological reasoning is currently written in Prolog and leverages the Maude rewriting system. As an example of some geo-Prolog code consider this… ‘Fault has top-layer Layer if Fault is a fault, Layer is a geological unit, Fault goes through Layer, and there does not exist a layer that is on top of Layer which Fault goes through’. The researchers claim that applying geological rewrite rules onto ‘proto-scenarios’ can deliver multi-scenario ‘beyond human capacity’.
© Oil IT Journal - all rights reserved.